OpenAI Downtime Monitor

Continuously polls OpenAI API endpoints and other LLM provider APIs at regular intervals (likely 30-60 second cadence) to detect availability status, recording binary up/down states and timestamps. Uses synthetic health check requests to measure actual endpoint responsiveness rather than relying on provider status pages, enabling detection of partial outages or regional degradation that official status pages may not reflect.

Measures response time for synthetic API requests to each monitored endpoint, recording latency metrics (likely p50, p95, p99 percentiles) and tracking latency trends over time. Aggregates latency data across multiple measurement points to identify performance degradation patterns, regional variations, or model-specific slowdowns that may not trigger uptime alerts but impact user experience.

Stores and visualizes historical uptime and latency data in time-series format, displaying trends through charts and status timelines. Likely maintains a rolling window of historical data (days to weeks) to show patterns, recurring issues, or seasonal variations in API availability and performance, enabling root cause analysis and capacity planning decisions.

Monitors a curated set of LLM providers and models beyond just OpenAI, including other major providers like Anthropic, Google, Cohere, and potentially others. Maintains a registry of monitored endpoints and models, allowing users to track uptime and latency across their entire LLM provider ecosystem from a single pane of glass without switching between multiple status pages.

Provides unrestricted public access to uptime and latency data through a web dashboard (status.portkey.ai) with no authentication or subscription required. Implements a freemium model where basic monitoring data is publicly available, potentially with premium features (alerts, webhooks, detailed analytics) available through paid tiers or integration with Portkey's broader platform.